BW video to colour conversion and simplify Iris

Could we have BW video to colour conversion? AVCLabs Video Enhancer AI is a cheaper alternative and has this feature.

Could you also possibly have a rethink of Iris face enhancer? Again AVCLabs Video Enhancer AI has this feature but it’s just a simple affair of using a slider to increase how much enhancement you want. No matter what I do with Iris, faces are ALWAYS too sharp and unnatural looking. So it’s a useless feature, in its current form IMHO.


AVCLabs is a pile of junk that is just a GUI to FREE software like BSRGAN and CODEFORMER and DEOLDIFY which they use for BW to colour.
Where is the licence file for DEOLDIFY ?
There isn’t one as per the MIT licence.
I wonder why ?
Oh yeah yet again they don’t want you to know thay are using others hard work in creating software and passing it off as their own work.

I agree that Iris is useless for most of the videos I do. Maybe if I was doing scene by scene it would work well.
BW to color is too much of a finicky idea. Sure you could do a ‘let the AI pick the colors’ approach. But for it to be actually professionally useful, the user would need to be able to fine-tune the colors applied to things. Like be presented with a report all the found objects, and the user can accept the suggested colors, or edit them. Anyway, I think we’re still a decade away from that.

Iris is literally the only reason I use this program. If it were ever removed, I would just ask for my money back.

@menditsa One feature AVCLabs Video AI has had for a long time that Topaz hasn’t been able to pull off yet is pausing/resuming processing. The interface is much simpler than TVAI and less intimidating to an average user. And it also lets you change the saturation and brightness of the videos.

@colin.feeney, I’ve requested an Iris HQ model that’s not so intense compared to the current models. My current workaround for instances where Iris goes a bit overboard is to either run Iris with all parameters set to zero (possibly a mildly positive deblur, no more than +15), or use Artemis HQ.

The problem is usually that the source video itself is too low-quality for the AI to handle. The TVAI models aren’t trained very well for the types of videos that most people are processing and are best suited for very-high quality content. If you can upscale the original video to a larger resolution first with either Artemis or Gaia (or a different program), Iris should run just fine.

Edit: now I am just using Iris as the preliminary enhancement. It turns out significantly better than Artemis. So my usage of the enhancement models has gone from 95% Iris to 100% Iris.

1 Like

Same here (except for my beta testing work where I have to test everything) :smiley:

1 Like